Margin and derivatives · standing brief, September 2026
What it costs to get out of a leveraged position
A perpetual position is opened on collateral and closed against whatever depth is resting at that moment. This brief reads eight order-book venues as exposure rather than as products.
The Margin Brief deskUpdated: Books polled to
Read firstSix venues by their audit recordThe formula, before the tables
Where the totals stand
Key figures in this issue
Summary of this issue
This publication covers one instrument. A perpetual future has no settlement date, it is carried on posted collateral, and the venue closes it when that collateral no longer covers the move against it. The leverage ladder, the fee schedule and the market count all describe the position while it is open. What decides the outcome is the exit, and the exit is priced by whatever book is resting at that moment.
So the weighting is deliberately lopsided. Depth within 10 basis points of the mid carries 30 of the 100 points and is the one criterion measured rather than read off a document; where the collateral sits carries 22, because posted margin is money handed over before anything goes wrong. The formula was fixed on 22 September 2026 and does not change from page to page: a page condition decides which venues appear, never how they are weighted. Nothing here is advice, and nothing on this site is cover, protection or a guarantee against a loss. Placement is paid for; the order of every table is not.
WeightingExit liquidity30Collateral and settlement22Cost per round trip20Loss record and audits14Contracts and asset classes9Getting a position on5the formula in full
What a position ties up, and what removes it
The less margin a position needs, the smaller the move that wipes it out. On EVEDEX, 200x on BTC holds to $50,000 of notional, so a 0.50% move consumes the margin; Aster's order book matches 200x only to 400 USDT. At 50x on Lighter, Extended, Paradex and dYdX the threshold is 2%, at Hyperliquid's 40x it is 2.5%.
Table 1Documented caps on Bitcoin, restated as the move they survive — caps read from each venue's own margin schedule
| Venue | Move that consumes the margin | Documented BTC cap | The notional it holds to |
|---|---|---|---|
| EVEDEX | 0.50% | 200x | $50,000 of exposure at the cap, $200,000 one rung down |
| Aster | 0.50% | 200x | 400 USDT of exposure at the cap, $800,000 one rung down |
| edgeX | 1.00% | 100x | $1 million of exposure at the cap, $3 million one rung down |
| ApeX Omni | 1.00% | 100x | one per cent posted to $100,000 of exposure |
| Extended | 2.00% | 50x | $4 million of exposure at the cap, $8 million one rung down |
| Lighter | 2.00% | 50x | two per cent posted, no exposure band published |
| Paradex | 2.00% | 50x | two per cent posted, capped at 300 BTC |
| dYdX | 2.00% | 50x | two per cent posted, no exposure band published |
| Hyperliquid | 2.50% | 40x | $150 million of exposure, halved above it |
| Aevo | 5.00% | 20x | five per cent posted, capped at $10 million |
Caps, and the notional each holds to, are read from each venue's own margin schedule or risk tiers on 18 September 2026. A cap almost never applies to the whole position: most of these venues step the multiple down as the notional grows, and the last column says where each stops. The middle column is arithmetic on the published cap and not a venue figure: at a hundred times, a move of one percent against the trade is worth the entire margin standing behind it. In practice the venue closes earlier, at its maintenance level, and funding and fees are charged on top of the move.
What one round trip costs on each book
Reading the figure
Both marks on a row are the same trade priced two ways. The open mark is what an opening and a closing order cost if both rest in the book and earn the maker rate; the filled mark is what they cost if both take liquidity. The gap between them is the part of the bill a trader still controls.
Figure 1Cost of a round trip — on $100,000 of exposure
both sides at the maker rateboth sides at the taker rate
Entry-tier published rates, read on 18 September 2026, applied to an opening and a closing trade. Bars are scaled to the dearest round trip here; the figure on the right is the taker cost. Cashback, rebates and volume tiers are ignored.
Standing of every venue this issue scores
Six criteria, weighted 30 · 22 · 20 · 14 · 9 · 5 and fixed on 22 September 2026. Scores run from 0 to 15, and the criterion column shows the raw points behind each total, each out of ten.
Table 2Every scored venue, in order of total — scores out of 15, criterion points out of ten
| # | Venue | Score of 15 | Points by criterion, each of ten | Documented cap on BTC | Median depth behind the exit | $1 million moves the price | Model |
|---|---|---|---|---|---|---|---|
| 1 | Lighter | 12.5 | liquidity 6.5 · custody 10.0 · cost 10.0 · security 7.0 · coverage 9.2 · access 8.0 | 50x | $16.9 million | 1.48 bps | Hybrid, ZK-proven |
| 2 | edgeX | 11.5 | liquidity 6.9 · custody 10.0 · cost 6.5 · security 7.0 · coverage 8.0 · access 8.0 | 100x | $29.3 million | 0.22 bps | Hybrid |
| 3 | Extended | 11.5 | liquidity 5.8 · custody 7.0 · cost 8.5 · security 10.0 · coverage 9.6 · access 8.0 | 50x | $7.8 million | 0.70 bps | Hybrid |
| 4 | Aster | 11.3 | liquidity 6.3 · custody 7.0 · cost 7.0 · security 10.0 · coverage 10.0 · access 8.0 | 200x | $14.8 million | 1.11 bps | Own layer 1, operator-run |
| 5 | Hyperliquid | 10.2 | liquidity 5.7 · custody 7.0 · cost 6.5 · security 7.0 · coverage 9.6 · access 8.0 | 40x | $7.4 million | 0.72 bps | On-chain order book |
| 6 | EVEDEX | 9.7 | liquidity 6.5 · custody 5.0 · cost 6.5 · security 7.0 · coverage 7.8 · access 8.0 | 200x | $18.4 million | 2.75 bps | Hybrid |
| 7 | Paradex | 9.0 | liquidity 0.0 · custody 7.0 · cost 10.0 · security 10.0 · coverage 7.0 · access 8.0 | 50x | $55.00 | not filled | Hybrid, Starknet appchain |
| 8 | Aevo | 7.7 | liquidity 4.3 · custody 7.0 · cost 3.0 · security 4.0 · coverage 8.5 · access 8.0 | 20x | $1.4 million | 7.63 bps | Hybrid |
The bar beneath each total measures it against the highest on the list, and nothing about the row is decided by hand. Criterion points are shown before the weighting, so a venue can sit high overall and still carry one weak column. The documented cap, the median depth behind the exit and what a million-dollar order moves are the three figures this file leads on; the cap is the venue's own, read on 18 September 2026, and the other two are our readings of its public book. The rubric behind every criterion is set out on how we rate.
What can be held there, and when
An account is worth only the exposures it can carry. These are the classes each venue lists beyond crypto, the number of perpetual markets its own interface reports, and whether the book ever shuts.
Table 3Asset classes, market counts and trading hours — as the venue's own interface reported them on 18 September 2026
| Venue | Asset classes | Perpetual markets | Trading hours |
|---|---|---|---|
| Lighter | crypto, stocks, FX, commodities, indices, pre-IPO, bonds | 214 | not stated in the documentation we read |
| edgeX | crypto, stocks, commodities, FX | 172 | not stated in the documentation we read |
| Extended | crypto, stocks, FX, commodities, indices, pre-IPO | 325 | not stated in the documentation we read |
| Aster | crypto, stocks, FX, commodities, indices, pre-IPO | 578 | not stated in the documentation we read |
| Hyperliquid | crypto, stocks, FX, commodities, indices | 324 | not stated in the documentation we read |
| EVEDEX | crypto, commodities, FX, stocks, indices, pre-IPO | 52 | All markets 24/7, shares and commodities included |
| Paradex | crypto, stocks, commodities, indices | 63 | not stated in the documentation we read |
| Aevo | crypto, stocks, commodities, FX, pre-IPO | 102 | not stated in the documentation we read |
A venue that lists a share, an index or a currency is quoting a perpetual contract on that price. Holding one gives the trader no claim on the asset underneath it, no dividend and no vote, and the contract settles in the collateral the venue accepts.
What is asked before a position can be opened
What a venue asks for before it lets an order through — how an account is opened, and what identity documents it wants — decides whether an account can exist at all. These are the first rows to read against the terms that apply where the trader lives.
Table 4Admission: what opens an account, and what papers are asked for — from each venue's own terms and help pages
| Venue | What opens an account | Papers asked for |
|---|---|---|
| Lighter | Wallet signature | None in the documentation we read |
| edgeX | Wallet or MPC login | None in the user terms we read |
| Extended | Wallet signature, EVM or Starknet | Terms reserve the right to deny access after a KYT risk assessment |
| Aster | Wallet signature or email | Terms allow limited identity verification procedures |
| Hyperliquid | Wallet signature or email code | None in the onboarding documentation |
| EVEDEX | Wallet signature | None to trade; on-chain AML screening on deposits |
| Paradex | Wallet signature; account derived from the address | None in the account documentation |
| Aevo | Wallet connect | Terms: transactions only through your own wallet service |
Taken from the terms and help pages each venue published on 18 September 2026. Admission rules are amended without notice and a venue may refuse an applicant on grounds it has never printed, so the terms in force on the day of the deposit are the ones that count.
Why the exit carries more weight than the fee
Terms used in this file
Margin is the collateral a trader puts up and leaves with the venue while a trade is open. Depth within ten basis points is the money resting close enough to the mid price to fill an order without moving it far. Price impact is what a market order of a stated size would have paid against the book at the moment it was read.
The two readings sit at opposite ends of the same measurement. Across the polls of 21 September 2026 and 22 September 2026 the edgeX BTC book held a median $29.3 million within 10 basis points of the mid, and a market order of $100,000 moved the price 0.02 bps. The Paradex book held a median $55.00, and the same order moved it 162.91 bps. Paradex charges nothing to retail orders from its own interface; edgeX charges 0.040% maker / 0.045% taker. A trader choosing on the fee schedule alone would have paid nothing to enter and more than a percent to leave.
A fee is known and the same size on the way in and on the way out. Slippage on the exit is not known, it is largest exactly when the position is being closed against the trader, and at ten times leverage it is charged against a tenth as much collateral.
What the collateral criterion is actually asking
Standing of this file
Venues pay to appear here. What they cannot buy is a position: the order of every table is produced by one published formula from figures that carry the date they were taken, and a venue that scores badly keeps its low score.
Three questions, all answerable from documents. How often does the venue write trades or positions to a chain, so that a record exists outside its own database. Is there a documented way to withdraw that does not need the operator. Were the contracts holding deposits audited by somebody else.
The answers separate this group more sharply than the fees do. Lighter documents a priority exit queue on Ethereum and a mode that lets users withdraw if the sequencer stops answering, and edgeX documents a withdrawal that bypasses the operator. Hyperliquid, Extended, Paradex and Aevo record every trade on a chain but document no such exit. On EVEDEX the matching sits off-chain and Arbitrum receives position data once enough of it has changed, which is a record in batches rather than trade by trade, and no operator-free withdrawal is documented there either. None of this predicts conduct; it is a list of what exists in writing.
Particulars of this issue
- Prepared by
- The Margin Brief desk, which keeps this file and answers for every figure in it.
- Subject
- A standing brief on what it costs to leave a leveraged position: the depth an exit is sold into, who is holding the collateral until then, and the charge on both fills. The depth is measured here; the rest is read off each venue's own paperwork, with the date of the reading beside it.
- Scope
- Venues that run a central limit order book for perpetual futures, whether the book is matched off-chain or held on a chain, and that this brief has taken into its standing coverage: eight of the sixteen on file, deliberately including the thinnest books rather than dropping them. Coverage is an editorial decision, closed with the weights and listed on this page; it is not part of the arithmetic. ApeX Omni and dYdX are on file and outside it in this edition, read and quoted wherever they bear on a point. Oracle-priced and pool-priced venues are listed with their facts but left unscored, because the depth criterion measures a book they do not have.
- Venues on file
- 16 venues are held on file and 8 of them carry a score; the remainder are listed with their published facts and left unrated.
- Basis of figures
- Venue documents and terms read on 18 September 2026. Depth, spread and impact are read off the public books over 122 polls between 21 September 2026 and 22 September 2026, by the method on how we rate. Open interest and volume over 24 hours is declared by the venue and listed by CoinMarketCap on 22 September 2026.
- Rating scale
- Totals run from 0 to 15 and are carried to 1 decimal place. The 6 criteria and their weights were fixed on 22 September 2026 and have not moved since.
- Standing
- Current issue. It is replaced only by a later issue at the same address, and a date moves only when the text does.
Issued by The Margin Brief desk. Amendments: editorial@insurancerefocused.com. Updated: .